|Publication number||US5606505 A|
|Application number||US 08/451,780|
|Publication date||Feb 25, 1997|
|Filing date||May 26, 1995|
|Priority date||Sep 17, 1993|
|Also published as||DE69411021D1, DE69411021T2, EP0719429A1, EP0719429B1, WO1995008144A1|
|Publication number||08451780, 451780, US 5606505 A, US 5606505A, US-A-5606505, US5606505 A, US5606505A|
|Inventors||Joseph A. Smith, Erik A. Ringnes, Victor F. Tarbutton|
|Original Assignee||Honeywell Inc.|
|Export Citation||BiBTeX, EndNote, RefMan|
|Patent Citations (6), Non-Patent Citations (4), Referenced by (31), Classifications (6), Legal Events (4)|
|External Links: USPTO, USPTO Assignment, Espacenet|
Γ=0 when M≦Mcr
Γ=15(M-Mcr)2 tan[(M-Mcr)2 (CL 3 +0.1)λ] when M>Mcr.
This application is a continuation-in-part, of application Ser. No. 08/122,214, filed Sep. 17, 1993 now abandoned.
The present invention relates to flight management systems. In particular, the present invention pertains to a method of estimating, and subsequently predicting, aircraft performance.
In the past, a "rule of thumb" method was utilized in predicting aircraft performance characteristics. With this technique, the pilot, based on previous experience, makes a reasonable estimate of the aircraft performance characteristics. Other past aircraft performance prediction systems include a custom performance computer which utilized "average parameters" for a predetermined model of an aircraft. This average parameter method is useful for only one aircraft type, and must be repeated for each new type of aircraft. These methods and systems have other inaccuracies associated therewith. For example, the rule of thumb method is very inaccurate since it is difficult to compensate for temperature of the air, weight of the aircraft, etc. Also, this method usually adds to the pilot workload at a time when the pilot is already very busy. The custom performance computer method, in addition to the average parameter disadvantage, does not account for manufacturing tolerances, a degradation of parameters resulting from age and usage, nor for different pilot techniques.
U.S. Pat. No. 5,070,458 to Gilmore, et al., entitled "Method of Analyzing and Predicting both Airplane and Engine Performance Characteristics," issued Dec. 3, 1991, describes a method which overcomes many of these disadvantages. This method makes performance predictions for an individual aircraft and engine using parameters which are learned from "flight to flight." The data, which is used by this method for predicting performance characteristics, is initialized with reasonable values of specific performance parameters. This method includes adjustment or updating of the specific performance parameters resulting from each flight of the aircraft. Thus, the performance characteristics of a given aircraft are learned from flight to flight for use in future flights of the given aircraft. The learned parameter technique thereby adjusts to changes in the aircraft due to aging, is tailored to a specific aircraft, and accounts for manufacturing tolerances. However, in Gilmore, et al., the modeling utilized and the computation of terms thereof are inefficient. In addition, Gilmore, et al. uses a model which separates thrust and drag terms of the model when making predictions and is somewhat limited in the number of outputs produced. For example, Gilmore does not provide for predictions of long range cruise speed or optimum altitude, which are important for achieving the best possible fuel efficiency.
In view of prior estimation and prediction systems, a need is apparent for a prediction system which provides an enhanced set of outputs for use by the pilot. In addition, the outputs, which may include outputs previously computed by prior systems, should be provided with greater speed and/or improved accuracy.
The present invention provides an estimation and prediction method which provides an enhanced set of outputs which are provided with greater speed and/or accuracy. An estimating method for use in predicting performance characteristics of an aircraft in accordance with the present invention includes modeling the aircraft with at least one series expansion using aircraft specific data, such as wing reference area, and input parameters, such as current weight or altitude, which define the performance characteristics. Coefficients of the at least one series expansion are learned over at least one flight so that the learned coefficients can be utilized for predicting performance characteristics of subsequent flights. A thrust and drag relationship may be modeled with a first series expansion and climb fuelflow of the aircraft may be modeled with a second series expansion. A first and second filter are utilized for estimating the coefficients of the first and second series expansions, respectively.
In another method of the present invention, a method of predicting the performance characteristics of an aircraft includes modeling the aircraft with at least one mathematical model using aircraft specific data and input parameters which define the performance characteristics. Coefficients of the mathematical model are learned during at least one flight and cruise performance characteristics are predicted as a function of the learned coefficients of the mathematical model. Long range cruise speed, maximum cruise speed and optimum altitude at both such speeds are predicted. These performance characteristics are computed using the mathematical model and a given set of input parameters, e.g. long range cruise speed at a given weight and altitude. These computations are optimized by differentiating functions derived from the mathemtical model with respect to each independent input parameter. In another method of the present invention, parameters such as total fuel burned for a given flightplan are computed with minimum iterations utilizing closed-form integration. The ability to differentiate the mathematical model is necessary to such integration.
An additional embodiment of the present invention includes a method of predicting performance characteristics of an aircraft including modeling thrust-minus-drag of an aircraft with at least one mathematical model using aircraft specific data and input parameters which define the performance characteristics. Coefficients of the at least one mathematical model are learned. Performance characteristics are predicted as a function of the learned coefficients representative of a combined and non-independent thrust-minus-drag relationship. The certain performance characteristics are a function only of the non-independent or combination thrust-minus-drag relationship, and not of thrust and drag independently. These predictions are of considerably greater accuracy than those depending on explicit knowledge of thrust and drag separately.
FIG. 1 shows a general block diagram of the method of airplane performance estimation and prediction in accordance with the present invention.
FIG. 2 shows a block diagram of the signal conditioning block of the learning portion of the method of aircraft performance estimation and prediction method of FIG. 1.
FIG. 3 shows a block diagram of the performance estimation block of the learning portion of the aircraft performance estimation and prediction method of FIG. 1.
FIG. 4 is a chart showing an aircraft flight.
FIGS. 5A-5C are graphs showing variations in key parameters as a function of altitude.
FIG. 6 is a graph showing short trip limited altitude.
FIG. 7, including FIGS. 7A-7C, show method blocks for determining speed in accordance with the present invention when altitude is fixed.
FIG. 8, including FIGS. 8A-8C, show method blocks for determining maximum and optimum altitude for manual, long range cruise, and maximum speed in accordance with the present invention.
FIG. 9 is a graph showing a specific range as a function of speed.
FIG. 10 is a graph utilized for determining maximum speed.
FIGS. 11A and 11B are graphs showing determination of optimum and maximum altitude of manual speed.
FIG. 12 is a contour graph showing thrust-minus-drag as a function of altitude and true airspeed.
FIG. 13 shows a specific range surface defined for determining long range cruise speed.
FIG. 14 shows a thrust-minus-drag surface defined for determining long range cruise speed.
FIG. 15 illustrates the invention embodied in a flight management system in communication with various aircraft sensing, control, and display systems.
The method of aircraft performance estimation and prediction 10 shall be described in general with reference to FIG. 1. The aircraft performance estimation and prediction method 10 includes two major portions, a learning portion 12 and a predicting portion 14. The learning portion 12 includes modeling a thrust-minus-drag relationship of an aircraft with a Taylor series expansion utilizing signal input parameters 26, such as current aircraft weight and altitude, and aircraft specific data, such as wing reference area, stored in non-volatile memory 32. Additionally, fuel flow at the climbing rate is modeled with another Taylor series expansion utilizing signal input parameters 26 and aircraft specific data. Performance estimation 16 receives conditioned signal inputs 28, and preferably utilizing Kalman filter techniques, estimates the coefficients of the Taylor series expansions of both the modeled thrust-minus-drag relationship and climb fuel flow over the course of multiple flights. The learning portion 12 yields performance estimate outputs 30 including fuel flow estimate vector xff, thrust-minus-drag estimate vector x, and covariance matrices P and Pff, of which P provides data for determining how accurate the system will predict aircraft performance characteristics utilizing figure of merit block 24 of prediction portion 14. The learned coefficients of the Taylor series expansions for the thrust-minus-drag relationship and fuel flow along with their respective covariance matrices are retained in non-volatile memory 32 and are therefore available during subsequent flights to refine the existing estimates utilizing learning portion 12 and to predict aircraft performance characteristics with climb predictions block 18, cruise predictions block 20 and descent predictions block 22.
The prediction portion 14 utilizes Newtons-method-related techniques to find certain points of interest on curves or surfaces representative of functions derived from the thrust-minus-drag relationship and fuelflow to be maximized or otherwise localized given particular constraints. Aircraft performance characteristics such as long range cruise speed, maximum speed, maximum and optimum altitudes and combinations thereof are examples of characteristics which are computed in this manner. Newtons-method-related techniques rely heavily on the ability to differentiate various functions of interest. This ability in turn depends ultimately on the ability to differentiate the thrust-minus-drag function and fuelflow which are expressed as Taylor series expansions whose coefficients are provided directly by performance estimation 16 and whose derivatives are readily available.
With the learned coefficients of the Taylor series expansions and respective covariance matrices retained in non-volatile memory 32, the data can be downloaded and utilized for initializing or "seeding" a new example aircraft of similar type. For example, a first aircraft can be utilized in the learning process, while the data learned can be used to seed the learning process of a second similar type aircraft or used for predictions of the second aircraft characteristics. In addition, with the coefficients and covariance matrices in non-volatile memory, the pilot is allowed to save the data if a box of the system is pulled for any reason. It can then be reloaded when desired mitigating the need to completely restart the estimation process.
With reference to FIGS. 1-14, the present invention shall be described in further detail. The non-volatile memory 32 includes an aircraft data file which is used for performance predictions once sufficient learning has taken place and also for continuing the learning process. Sufficient learning in the learning portion 12 has taken place to begin the prediction portion 14 usually after about 3-4 flights. The aircraft data file is divided into two parts. The first file part contains data that is independent of learning, but still aircraft specific. This aircraft specific data includes the following:
aircraft tail number
ceiling altitude (feet), hceil
default climb constant calibrated air speed (KCAS)
default climb Mach
default descent KCAS
default descent Mach
takeoff fuel allowance (lbs)
landing fuel allowance (lbs)
basic operating flight weight (BOW) (lbs)
wing platform area (ft**2), S
maximum operating airspeed, Vmo (KCAS)
maximum operating mach number, Mmo
maximum lift coefficient, CL.sbsb.ref
critical mach number, Mcr
The second file part of the aircraft data file contains the learned coefficients of both the Taylor series expansions and their respective covariance matrices. The learned data of the second file part of the aircraft data file includes:
11 coefficients for the thrust-minus-drag Kalman filter
5 coefficients for the rated-fuelflow Kalman filter
the upper triangle of the eleven-by-eleven covariance matrix including the diagonal elements describing the thrust-minus-drag filter (a total of 66 values)
the upper triangle of the five-by-five covariance matrix including the diagonal elements describing the rated-fuelflow filter (a total of 15 values).
As indicated above, the aircraft is modeled by use of Taylor series expansions of thrust-minus-drag and rated fuelflow about a given reference point. For the remainder of the present description, given a Taylor series expansion of the function z=f(x,y) of the form: ##EQU1## then fo, fx.sbsb.o, fy.sbsb.o, etc. will be referred to as coefficients of the series,
Δx, Δy (or x and y with the "--xo " understood), etc. will be referred to as parameters of the series, and
fx.sbsb.o Δx, etc. are terms of the series.
The terms "coefficients" and "parameters" are also used in the context of a general mathematical model.
The Taylor series expansion for the thrust portion T of the thrust-minus-drag relationship is expressed as: ##EQU2## where M is Mach number, δ is pressure ratio, Wfc is corrected fuelflow defined further below, hp is pressure altitude, and |o indicates the preceding term is to be evaluated at the midpoints Mo, Wfc.sbsb.o and hpo which are also further defined below.
The drag portion D of the thrust-minus-drag relationship is expressed as:
D=148δM2 SCD =C1 δM2 CD
S=wing reference area
CD =coefficient of drag
The coefficient of drag CD is modeled as:
CD =Cd.sbsb.i (1+Γ)
CD.sbsb.i =coefficient of drag-incompressible flow
Γ=compressibility correction as described below.
The incompressible drag coefficient CD.sbsb.i is modeled as:
CD.sbsb.i =CD.sbsb.o +CD.sbsb.l CL +kCL 2
CL =coefficient of lift
CD.sbsb.o, CD.sbsb.l, & k are coefficients to be learned.
Since lift L can be written as Cl δM2 CL and is equal to weight W in flight, drag D can be expressed as: ##EQU3##
The compressibility correction term Γ is expressed as:
Γ=0 when M≦Mcr
Γ=15(M-Mcr)2 tan[(M-Mcr)2)(CL 3 +0.1)λ] when M>Mcr
Mcr =critical mach number
λ=coefficient to be learned
Combining the expressions for thrust and drag, the thrust-minus-drag relationship T-D is modeled as follows: ##EQU4## Taken collectively, the coefficients x1, x2, etc. can be expressed in vector form as x=[x1, . . . x11 ]T. The performance estimation portion 16 of the present invention develops estimates of these true coefficients; the estimates denoted by x1, x2, etc. or by the vector x.
The 11-term Taylor series expansion allows more degrees of freedom than did the prior art, resulting in the modeling of significant higher order effects which yield better results. Also, the altitude hp becomes an intrinsic part of the modeling in the present invention, unlike other methods of estimation and prediction which utilize an "after the fact" curve fit of altitude with other parameters.
The reference point about which the Taylor series is expanded is chosen for best fit over typical operating conditions. In the preferred embodiment, it is defined to be Mo =0.60 mach, hpo =30,000 feet, and Wfc o =7,400 lbs/hr, thus:
ΔM=M-0.60; Δh=hp -30,000; ΔWfc =Wfc -7,400
The rated fuelflow, also modeled by a Taylor series expansion, is expressed as: ##EQU5## where TDISA =T-TISA ; TISA being standard temperature, or may be expressed as:
Wfr =xff1 +xff2 ΔM+xff3 Δh+xff4 TDISA +xff5 ΔhTDISA
The rated fuelflow Wfr is modeled by the Taylor series expansion about the same reference point as thrust-minus-drag is modeled. Taken collectively, the coefficients xff1. . . xff5 can be expressed in vector form as xff =[xff1 . . . xff5 ]T. The performance estimation portion 16 of the present invention develops estimates of the coefficients; the estimates denoted by xff1 . . . xff5 or by the vector xff.
With specific reference to FIGS. 2 and 3, the learning portion 12 of the present invention shall be described in further detail. Input signal conditioning is provided by integrator 34, corrected fuelflow block 36, atmospheric model 38 and two steady-state Kalman filters, altitude rate filter 40 and acceleration rate filter 42. Input signal conditioning block 15, FIG. 2, receives the signal parameter inputs 26 including fuelflow Wf, mach number M, temperature T, pressure altitude hp, pitch θ, roll φ, sensed longitudinal acceleration aLON, sensed normal acceleration aNOR and true air speed VT. The inputs Wf, M, VT, and hp are also inputs to performance estimation block 16.
Fuelflow Wf is applied to integrator 34 along with initial aircraft weight Wo, determined from specific aircraft data including BOW and other pilot input as known to one skilled in the art, to provide an estimated weight W as fuel is utilized. Corrected fuelflow block 36 for output of corrected fuelflow Wfc is provided with inputs of fuelflow Wf, mach M, and temperature T. Corrected fuelflow Wfc is defined by: ##EQU7## where δ is the pressure ratio and θ is temperature ratio. Pressure ratio δ and temperature ratio θ are provided to corrected fuelflow block 36 from atmospheric model block 38 which has the inputs of temperature T and pressure altitude hp. The standard atmospheric relationships which determine these inputs to corrected fuelflow block 36 are well known and understood by those skilled in the art.
By means of integrator 34, corrected fuelflow block 36 and atmospheric model 38, conditioned inputs 28 including estimated gross weight W, corrected fuelflow Wfc, pressure ratio δ and the deviation from standard temperature TDISA are provided to performance estimation block 16.
The altitude rate filter 40, preferably a steady state Kalman filter, provides the estimated output of altitude rate hp. The general equation for a Kalman filter is written as
with z being the measurement vector, K being the gain matrix, M being a matrix derived from the gain and transition matrices, xk+1 being a (2,1) new state vector and xk being the previous state vector. The only input to the altitude rate filter 40 is the pressure altitude hp ; z vector for the filter is of dimension (1,1). The K and M matricies are given by: ##EQU8## The (2,1) new state vector xk+1, consists of estimates of altitude and altitude rate. The desired altitude rate output hp is the second element of this new state vector.
The acceleration filter 42, also preferably a steady-state Kalman filter and generalized in the same manner as altitude rate filter 40, provides a smooth estimated output rate of airspeed VT. However, simply computing the time rate of change of VT from the airspeed input VT is insufficient. Thus, the filter also uses inertial accelerations. The acceleration in the stability reference frame is given by:
ax =cos a(aLON =sinθ)+sin a(-aNOR +cosφcosθ)
where the angle of attack a is computed from: ##EQU9## with θ being pitch angle.
The measurement vector z of this filter is: ##EQU10##
The K and M matricies for this filter are given by: ##EQU11##
The desired output of rate of airspeed VT is the second element of the new state vector for this filter.
As described above, the conditioned inputs 28 and several signal inputs 26 are provided to performance estimation block 16. They include gross weight estimate W, corrrected fuelflow Wfc, pressure ratio δ pressure altitude hp, deviation from standard temperature TDISA, fuelflow Wf, altitude rate hp, airspeed rate VT, and true air speed VT. Performance estimation block 16, FIG. 3, utilizes two Kalman filters to learn the coefficients of the Taylor series expansions modeling thrust-minus-drag and rated fuelflow as previously described. Performance Kalman filter 48 is utilized to learn the coefficients of the thrust-minus-drag relationship and fuelflow rating Kalman filter 52 is utilized to learn the coefficients of the Taylor series expansion modeling rated fuelflow. The performance Kalman filter 48 is an extended Kalman filter necessary for estimating coefficients of a non-linear function.
Kalman filtering techniques are known to those skilled in the art. This conventional approach to filtering involves propagation of a state estimate vector and a covariance matrix from stage to stage. It should be readily apparent to one skilled in the art that any appropriate filtering technique may be used in this performance estimation portion of the invention to provide the process of learning coefficients of a mathematical model. Such filtering could also be utilized where filters are used in signal conditioning block 15. Likewise, it is recognized that neural networks could be used in lieu of the series expansion modeling technique.
There are five recursive equations that make up a Kalman filter used in the present invention for filters 48,52. Preferably, they are run on a 10 second cycle as long as qualifying measurements are available. The state vector is expressed as x. The covariance matrix P describes the degree of uncertainty in x. The process is assumed to be subject to zero-mean, Guassian white noise described by a Q matrix which is diagonal. The measurement vector is z and its associated covariance matrix is given by R.
Since, in the cases of the performance Kalman filter 48 and the fuelflow rating Kalman filter 52, essentially steady-state coefficients are being estimated, the respective state transition matrices Φ become identity I, and the standard Kalman filter equations reduce as follows:
x- =Φxk =xk
P- k+1 =ΦPk ΦT +Q=Pk +Q
gain computation: ##EQU12## since R is scalar in these cases. update equations:
Pk =(I-Kk H)P- k
xk =xk +Kk (zk -Hx- k)=(I-Kk H)xk - +Kk zk
In accordance with the general description of Kalman filters above, the T-D state estimate vector x is of dimension (11,1) including the coefficients of the T-D Taylor series expansion as discussed previously. The state vector x is given by:
x=[x1 . . . x11 ]T
Therefore, the update vector H is of dimension (1,11) and consists of the parameters of the T-D series. The update vector H is computed as shown by block 46, FIG. 3, as a function of Mach number M, pressure ratio δ, weight W, corrected fuelflow Wfc and altitude hp. The update vector H is provided to performance Kalman filter 48. The first seven elements of update vector H are given by the thrust update vector HT and the last four are given in the drag update vector HD.
The thrust update vector HT is as follows:
HT =δ[1, ΔM, ΔWfc, Δh, (ΔM)2, ΔMΔWfc, (ΔWfc)2 ]
The drag update vector HD is Mach number dependent. The incompressible HD or HD.sbsb.i is: ##EQU13## where C1 is (1481)(S).
The compressible HD or HD.sbsb.c is only computed when the Mach number M is above the critical Mach number Mc.sbsb.r and is given by: ##EQU14## and where the summation run is from i of 8 to 10. Finally, when
M<Mcr then HD =Hd.sbsb.i
M≧Mcr then HD =HD.sbsb.i (1+Γ)+HD.sbsb.c
Because of the greatly differing ranges of the variables in update vector H, it is normalized by a matrix N. This normalized update vector H' is then used in the Kalman filter equations. This vector H' is defined as:
where the normalization matrix, N, is a diagonal matrix of dimension (11,11). The diagonal elements are defined as:
As shown by block 44, the measurement vector, z, of dimension (1,1) and representative of thrust-minus-drag T-D is computed from the gross weight estimate W and the inputs of true airspeed VT, altitude rate hp and rate of airspeed VT as follows: ##EQU15## where g is the gravitational constant of 32.174 ft/sec2, f1 is a knots to fi/sec conversion factor of 1.6878065, and f2 is a ft/sec to knots conversion factor of 0.5924850.
With the associated error covariance R (1,1) of measurement vector z set to a predetermined vector and the diagonal elements in Q (11,11) set to a predetermined vector, x', the normalized form of x, can be computed. Not-normalized x is then computed from the Kalman filter output of x and N as follows:
In order to take a measurement of the quantity thrust-minus-drag T-D, an estimate of current acceleration must be available. In a cruise condition of the aircraft, current acceleration is zero and a measurement is not taken unless this condition is ensured. However, in climb or descent of aircraft acceleration is not zero. In prior art systems, acceleration was estimated from a planned speed schedule; however, this schedule could be violated in actual flight. The present invention bases its acceleration estimate on currently estimated values of rate of change of true air speed VT and altitude rate hp utilizing inertial acceleration components. Also, measurements are processed only when estimated acceleration is within acceptable limits. The following conditions determine when the thrust-minus-drag equation is valid and consequently when to run the Kalman filter for the learning process:
The airplane must be 5,000 ft above the takeoff runway.
The bank angle must be less than 5 degrees.
The rate of change in pitch must be less than 0.1 degree per second.
The Mach number must be greater than 0.40.
After x has been learned, the coefficients thereof can be used in the prediction portion 14 of performance estimation and prediction method 10 in subsequent aircraft flights.
The Kalman filtering approach propagates the covariance matrix rather than information which is the inverse of the covariance matrix. This leads to a much more computationally efficient form of filter. With use of normalization and double-precision arithmetic more maintainable software is achieved without compromise of accuracy.
Although the individual terms of the mathematical model can be separated as to which apply to thrust and which apply to drag, the estimate of the combined relationship of thrust-minus-drag (T-D) is more accurate than individual estimates of thrust and drag as they are not independent. Greater accuracy can be achieved in prediction of characteristics by expressing most of the computations required for prediction in a form that does not depend on such separation.
With regard to fuelflow estimation, the fuelflow rating Kalman filter 52 is run when the following filter conditions are satisfied. The core filter equations for the filter 52 have already been discussed with regard to performance Kalman filter 48. The conditions include:
The aircraft must be more than 1,500 ft above the takeoff runway.
The aircraft vertical speed must be greater than a minimum vertical speed Vmin defined as (units of ft/min): ##EQU16## The aircraft must be controlling to airspeed while climbing and at the same time be 1,000 ft below any preselected level off altitudes.
The fuelflow must not have changed more than 400 lbs/hour since the last 10 second sample of the filter. If it has, a 30 second timer is started before learning is again enabled.
The fuelflow state estimate vector xff is of dimension (5,1) and is defined in accordance with the partial derivatives of the Taylor series expansion modeling fuelflow. The state vector xff is given the subscript ff to distinguish it from the thrust-minus-drag state estimate vector and is given by:
xff =[xff1 . . . xff5 ]
The update vector Hff is computed as a function of Mach number M, pressure altitude hp and deviation from standard temperature TDISA as shown by Hff matrix computation block 50. Update vector Hff is of dimension (1,5) and is defined by:
Hff =[1, ΔM, Δhp, TDISA, TDISA Δhp ]
Because of the greatly differing ranges of the variables in Hff, it is normalized by matrix Nff resulting in normalized H'ff which is used in the Kalman filter techniques and is given as follows:
H'ff =Hff Nff
where Nff is a diagonal matrix of dimension (5,5); the diagonal elements thereof given by:
Nff =[1000, 300, 0.05, 10, 0.005]
The measurement vector zff (1,1) is the current unfiltered and uncorrected fuelflow Wf. With the associated error covariance Rff of zff set to a predetermined vector and the diagonal elements in Qff (5,5) set to predetermined values, x'ff, the normalized vector can be computed. Not-normalized xff is then computed from the Kalman filter output of x'ff and Nff as follows:
xff =Nff xff '
The vector xff can then be utilized by predicting portion 14 of aircraft performance estimation and prediction method 10.
With rated climb fuelflow Wfr being modeled as a function of Mach number M, altitude hp and deviation from standard temperature at altitude TDISA, significantly improved estimates of fuel consumed during climb are achieved. In addition, the results of this estimation are employed to predict characteristics such as maximum cruise speed.
The state estimate vectors xff and x and the covariance matrices Pff and P are part of the aircraft data file and stored in non-volatile memory 32. Each time a transition from airborne true to false is detected, an aircraft flight counter is incremented and the file updated with the latest values for such matrices. Only if a new aircraft data file is loaded is the data replaced; the learned data from one aircraft being capable of usage for seeding other aircraft. The initial values of xffo for the first flight in a new aircraft type are set to:
xffo =[400, 1000, -0.15, -25, 0.01]
The initial values of xo for the first flight on a new aircraft type are set to:
xo =[12000, 0, 2, 0, 6000, 1, 0, 0.02, -0.01, 0.1, 150],
and the Po and Pffo matrices are set to identity matrices.
Data from a first application of the method on a new aircraft type creates a data file to be used as a starting file or seeding file on other aircraft of the same type. The xffo matrix, the xo matrix, and the diagonal elements of Po and Pffo will remain unchanged; however, all non-diagonal elements in the Po and Pffo matrices are set to 0.
With a computer containing the present invention including an aircraft data file that is sufficiently learned, various predictions can now be accomplished via the predicting portion 14 of the performance estimation and prediction method 10. Such various predictions are made with regard to climb, cruise, and descent and as a function of modeled thrust-minus-drag and/or modeled rated-fuelflow.
FIG. 15 illustrates the invention embodied in a flight management system(FMS) in communication with various aircraft sensing, control, and display systems. In the preferred embodiment, learning portion 12 and predicting portion 14 are implemented in software running on a computer as an integral part of FMS 150. Non-volatile memory 32 is also a part of FMS 150. FMS 150 is in communication with various aircraft systems which supply input parameters 26 to learning portion 12 and predicting portion 14. Air data computer(ADC) 151 supplies mach(M), temperature(T), altitude(hp), and true airspeed(VT) data. Inertial reference system(IRS) 152 supplies pitch(θ), roll(φ), longitudinal acceleration (aLON), and normal acceleration (aNOR) data. Fuel sensor 153 supplies fuel flow (Wf) data. Each of these aircraft systems, and their equivalents, are well known in the art and need not be further described.
In one embodiment, performance predictions 154 are communicated to a control display unit(CDU) 155. The aircraft pilot or operator uses the displayed performance predictions to control the aircraft to the optimal flight profile. It is envisioned that any type of display device may be used to communicate performance predictions to the pilot including, but not limited to, cathode ray tubes, flat panel displays, heads up displays(HUDs) and the like. In alternate embodiments, performance predictions are communicated to an autopilot(A/P) 156 and an autothrottle(A/T) 157 which automatically control the aircraft to the optimal flight profile. Those skilled in the art understand that performance predictions would normally be transformed into a suitable format prior to communicating performance predictions 154 to CDU 155, A/P 156, or A/T 157.
An aircraft flight is divided into three phases: climb phase, cruise phase and descent phase, FIG. 4. The climb phase occurs from the origin of the flight to top of climb (TOC). The cruise phase extends from TOC to top of descent (TOD). Finally, the descent phase begins at TOD and continues through destination. Predictions for these phases of the flight are made via the method blocks: climb prediction 18, cruise predictions 20 and descent predictions 22.
Climb predictions 18 provide the following information: time enroute, t, fuel remaining Fr, altitude hp, airspeed VT, and groundspeed VG at each waypoint in a flightplan. Also, the top of climb predictions including the distance-to XTOC and fuel remaining at top of climb FTOC are determined. The prediction methods involve extensive use of first order Taylor Series expansions. The quantities climb angle γ, rated fuelflow Wfr, and true airspeed VT, are all assumed to vary linearly with altitude as shown in FIGS. 5A-5C. They are written as: ##EQU17## where ACF is a standard acceleration factor known to those skilled in the art, W is the aircraft gross weight, VG is groundspeed, and subscript o means the quantity evaluated at a reference altitude.
The climb angle γ is referenced to earth fixed coordinates and is therefore corrected for the effect of winds. The rate of change of γair with respect to altitude hp is written: ##EQU18## By assuming that the wind increases with altitude, the rate of change of climb angle γ with altitude hp is also corrected for the effects of wind by: ##EQU19##
The rate of change of rated fuelflow with respect to altitude ##EQU20## is readily available from the rated fuelflow Wfr equation previously discussed. Similarly, the rate of change of thrust-minus-drag with respect to altitude ##EQU21## is available from the thrust-minus-drag T-D equation previously discussed. The rate of change of true airspeed with altitude ##EQU22## can be found by assuming either a constant calibrated airspeed or a constant mach number and either a constant lapse rate or a constant temperature.
Distance x, time t, and fuelburn, F, are now written as integrals with respect to altitude hp in forms which allow the solutions to be found using standard integral tables. The integrals are expressed as: ##EQU23##
The climb integration is typically subdivided into three steps. These steps are: a lower step where it is assumed that the aircraft flies at a constant calibrated airspeed, a middle step where the aircraft flies at a constant Mach number but is below the tropopause, and finally an upper step where the aircraft is flying at a constant Mach number but is above the tropopause. The step altitudes are determined by a specific aircraft's climb KCAS/Mach defined in the aircraft data file or entered by the pilot. For the ##EQU24## and for the two other steps, ##EQU25## can be determined assuming a standard lapse rate. The cruise altitude chosen, either an optimum altitude or a pilot entered altitude, may be set low enough to preclude one or more of these steps. The starting altitude for climb predictions is always the aircraft altitude. Thus, as the aircraft climbs, the number of climb steps are reduced. Once in cruise, no climb predictions are made. The reference altitude for each step, where the various quantities are evaluated, is always chosen to be the step's mid-altitude.
Distance x as a function of altitude hp which is used for TOC placement, becomes: ##EQU26## where hc is the altitude at the end of the step, hb is the altitude at the beginning of the step, and ##EQU27## By rewriting this equation, altitude hp as a function of distance x, can be found for use in making altitude predictions at waypoints, or in this case at the end of a step. ##EQU28##
The solutions to time t and fuelburn F are subdivided into two parts, one when VT is constant and second when VT is changing with altitude hp. For constant VT : ##EQU29## For dVT /dhp ≠0 ##EQU30## With fuelburn F, time t, and distance determined for each step, values at top of climb such as xTOC and FTOC are simply the sum of all the individual steps.
Cruise predictions 20 provide estimates of time enroute tTOD and fuel remaining at TOD FTOD and distance to TOD xTOD. The description provided below for cruise predictions of total cruise fuelburn F and time t is described with respect to a single leg flight plan; however, it is equally applicable to a multiple leg flight plan in which case predictions are provided at each waypoint in cruise. In cruise, the altitude hp remains constant, but the aircraft's gross weight W might change significantly depending upon the length of the flight plan. Required fuelflow Wf decreases correspondingly as weight W decreases. In addition, the groundspeed VG can be influenced significantly by the winds, and to a much lesser extent by temperature changes. Furthermore, when flying either at the maximum speed Mmax or at the long range cruise speed MLRC, the mach number M (or the calibrated airspeed) will change with weight W. Like the climb prediction method, it is assumed that fuelflow Wf, airspeed VT, and winds vary linearly. This allows for use of first order Taylor series expansions, which again produce integrals that can be solved using standard techniques.
First, fuelflow Wf is known as a function of weight W from the thrust-minus-drag T-D equation by setting T-D equal to zero. The rate of change of fuelflow with respect to weight ##EQU31## can then also be found using standard methods for computing derivatives. Thus, fuelflow Wf can be written as: ##EQU32## where Wfo =initial fuelflow at TOC and Wo =initial gross weight at TOC from climb predictions block 18.
Cruise time as a function of weight can be written as the following integral: ##EQU33## This integral can be solved to yield time as a function of weight, but the useful form is weight or fuelburn as a function of time. Casting the equation in the form of fuelburn F as a function of time t, yields ##EQU34## Distance traveled x as a function of time is given by the following integral:
where VG =VT cos(λ)-WH with λ=drift angle and WH is the average component of the wind along the ground track.
VT is related to the Mach number M, assuming that the Mach number M changes linearly with time as follows: ##EQU35## where "a" is the speed of sound at the cruise altitude and for an average cruise leg temperature. An expression for the average drift angle is found by integrating over the crosswind, assuming again a linear change as expressed by: ##EQU36## where WC is the component of wind normal to the ground track with WC.sbsb.b =beginning WC and WC.sbsb.c ending WC.
Time then becomes the solution of a second order equation: ##EQU37##
Mach number M changes with time depending upon both temperature and wind at the beginning and the end of a cruise leg. The method employed to determine ##EQU38## is as follows. A first estimate of cruise time, assuming ##EQU39## is zero, is used to find time t. This allows computation of fuelburn F, and predicted Weight W, at the end of the leg or as in our case the end of the cruise phase. Then, the Mach number M at the end of the leg is computed at the first estimate of W. ##EQU40## can then be evaluated. If a constant Mach number cruise mode is chosen, as discussed later herein, this step is not necessary. With ##EQU41## computed, the time t and fuel estimates are then refined.
The effect of a changing wind WH can be significant because the aircraft will spend more time in the headwind than it will in the tailwind. The above estimates of time were determined using an average WH. When the wind changes significantly, a slowdown factor fs can be estimated by: ##EQU42## The original time estimate can now be multiplied by the slowdown factor fs to give the final time and fuelburn values.
Descent predictions 22 are performed in a similar manner to the climb predictions 18. For descent predictions block 22, which utilizes x, the inputs of a descent schedule, winds, and estimated cruise altitude hPcrz, two basic modes can be utilized. The first mode is called a path descent method where the ground referenced descent angle γd is constant. This method fixes the TOD placement regardless of winds predicted. The other method fixes the descent angle γd and consequently moves TOD based on predicted winds.
For short flight plans, the optimum altitude hp may be limited by how high the aircraft can go before a descent is necessary. The following rule described with reference to FIG. 6, is employed: the cruise phase should be at least 30 nautical miles, but no more than one quarter of the total flight plan distance 1/4xtot. A corrected distance xdist, can be constructed accounting for the origin and destination elevation and subtracting the cruise distance. The short trip limited altitude, hs, can then be estimated. It is assumed that the descent angle γd is constant and that the climb angle γc decreases linearly with altitude. Using the standard Sine Rule with a small angle approximation and the first order Taylor series expansion for the climb angle γc leads to the following second order equation for hs : ##EQU43## This equation can then be solved to yield hs which provides the short trip limited optimum cruise altitude.
Along with the predictions of fuelburn, time, and other performance characteristics such as climb angle, predictions for speed and/or altitude in various speed modes of cruise predictions block 20 as shown in FIGS. 7A-7C and FIGS. 8A-8C are performed utilizing Newton's method related techniques. Newton's method related techniques allow for determining certain points of interest on curves that represent functions to be maximized or otherwise localized given particular constraints. This leads to the ability to compute speed in various speed modes as shown in FIG. 7A-7C with altitude fixed. They include long range cruise speed, FIG. 7A; maximum speed, FIG. 7B; and maximum endurance speed, FIG. 7C. It also leads to the ability to compute altitude with speed fixed as in manual speed at optimal altitude, FIG. 8A. In addition, in higher dimension cases, Newton's method related techniques are used to find certain points of interest on surfaces that represent functions to be maximized, such as where speed and altitude are variable. Such modes include long range cruise speed at optimal altitude, FIG. 8B, and maximum speed at optimum altitude, FIG. 8C. In the case of long range cruise, speed and altitude are determined for providing best fuel mileage, or in other words, the speed which maximizes specific range SR is determined; specific range being defined as: ##EQU44## wherein VG is ground speed and Wf is fuelflow.
With mach M being determined as a function of fuelflow Wf while altitude and weight are fixed, and T-D is taken to be zero or some small positive constant, Newton's method related techniques shall be described for calculation of long range cruise speed MLRC, FIG. 7A. To describe calculation of long range cruise speed, a curve, FIG. 9, is shown wherein the x-axis is mach number M and the y-axis is specific range SR which is a function of mach number M and fuelflow Wf. Maximum range cruise mach Mmrc is the mach number for which SR (M) is a maximum. Long range cruise Mach MLRC is the point above Mmrc at which SR (M)=0.99SR (Mmrc). Mmrc is determined in accordance with the following steps with altitude hp and weight W being given parameters:
a) Choose an initial estimate, M1 : ##EQU45## where C1 =1481(S), W is gross weight and δ is atmospheric pressure ratio.
b) Compute the specific range SR and the derivative of specific range SR ' with respect to mach at this point: ##EQU46## The details of this computation are described further below. c) Choose a second initial estimate, M2 : ##EQU47## where Ms =mach step size=0.05 initially d) Compute SR.sbsb.2, and SR.sbsb.2 ' in the manner like explained below for SR.sbsb.1 and SR.sbsb.1 '.
e) Estimate the second derivative SR ": ##EQU48## f) Check if the second derivative is non-negative. If non-negative, the current "best fit" parabola has a minimum so no maximum will be found. This implies that the real SR function has an inflection point and the search region must be moved past the inflection point which is also away from the current minimum: ##EQU49## where MIA =mach "inflection adjustment"=0.2 If the second derivative is negative, the current best-fit parabola has a maximum which is "close" to the real Mmrc . This maximum is the point at which SR (M)=0: ##EQU50## Ensure that this new value for Mmrc is not farther than MIA from M1. If so, set Mmrc =M1 ħMIA whichever limits the movement of Mmrc without changing its direction.
g) Reduce the mach step size Ms =Ms /2.
h) If the absolute value of Mmrc -M1 >0.003, revise the initial estimate: M1 =Mmrc and repeat steps b)-g), otherwise go to step i).
i) Compute SR.sbsb.mrc =SR (Mmrc) and SR.sbsb.mrc '=SR '(Mmrc).
j) Estimate the second derivative at this final maximum: ##EQU51## k) Compute the long range cruise speed: ##EQU52##
The computation of specific range SR as for a given value of mach M, and of the first derivative, ##EQU53## step b) above, given specific values of pressure altitude hp, temperature deviation TDISA, weight W and current coefficients x are as follows:
ΔM=M-Mo and Δhp =hp -hpo
Mo =0.6 and hpo =30,000 ft
b) Compute temperature ratio θ and pressure ratio δ from hp and TDISA by standard atmospheric relationships.
c) Compute true airspeed VT : ##EQU54## where ao is the standard sea-level speed of sound=661.4748 kts. d) If M>Mcr, perform steps e)-g) to compute Γ, ##EQU55## otherwise set ##EQU56## and continue with step h). e) Compute coefficient of lift CL : ##EQU57## where C1 =1481(S) f) Compute two interim terms, ζ and ξ:
ζ=15(M-Mcr)2 ; ξ=(M-Mcr)2 (CL 3 +0.1)
g) Compute Γ, ##EQU58##
Γ=ζtan (ξx11) ##EQU59## h) Thrust-minus-drag T-D as a function of corrected fuelflow Wfc is written as a quadratic in ΔWfc :
T-D=aΔWfc 2 +bΔWfc +c=Trc
where Trc is residual climb thrust and is given by: ##EQU60## where hrc is taken (in this case) to be 100 fpm and the quadratic coefficients are computed as:
b=δ(x3 +x6 ΔM)
c=Trc +δ(x1 x2 ΔM+x4 Δhp x5 ΔM2)-hD
where ##EQU61## i) Find the solution to the quadratic: ##EQU62## if|a|>ε, otherwise ΔWfc =-c/b where ε is a small positive constant.
j) Find fuelflow for this mach number:
Wfc =ΔWfc +Wfc.sbsb.o,
where Wfc.sbsb.o =7400 lbs/hr. ##EQU63## k) Find groundspeed VG and specific range SR :
VG =VT -WH
where HH =headwind ##EQU64## l) Compute the partial derivatives of fuelflow with respect to mach and to corrected fuelflow: ##EQU65## m) Compute the total derivative of corrected fuelflow with respect to mach: ##EQU66## n) Compute the total derivative of corrected fuelflow with respect to mach: ##EQU67## o) Compute the partial derivative of specific range with respect to mach: ##EQU68## where Wc is crosswind p) Compute the total derivative of specific range with respect to mach at the desired mach: ##EQU69##
To determine maximum speed Mmax at fixed altitude, described with reference to FIG. 7B, the mach number M which drives T-D to Trc with fuelflow at the rated value (Wf =Wfr) is found. This is performed by "guessing" M and solving for Wf and ##EQU70## in a similar manner as done for the case of determining MLRC. Performing this at two initial values of M allows the computation of Mmax to drive Wf to equal Wfr as shown in FIG. 10.
Maximum endurance speed, Vme, FIG. 7C, is first computed as:
Vme.sbsb.o =Vref +10
where ##EQU71## and po =standard sea level air density=0.00237692 slugs/ft3.
CL.sbsb.ref =reference aircraft coefficient of lift-obtained from the specific aircraft data file.
Vme.sbsb.o is then checked for being thrust limited as follows: ##EQU72## where VTL, thrust-limited speed is taken from the mach number M which drives (T-D)|w.sbsb.fr to Trc as in the maximum speed case, the only difference being the smaller solution is chosen. In the special case where no solution exists, the M which maximizes (T-D)|w.sbsb.fr is used.
The prediction of manual speed mode, or optimal hpopt or maximum altitude hpmax for manual speed, FIG. 7A, includes three submodes: a KCAS only, a Mach only, or both KCAS and Mach specified. Single dimension searches of SR and T-D are used to find hpopt and hpmax altitude, respectively. In order to perform those searches, the ##EQU73## and ##EQU74## must include the ##EQU75## for the different speed submodes. These can be derived from standard expressions and using a standard lapse rate. The max altitude hpmax search estimates the place where (T-D)=0 as shown by FIG. 11A, whereas the opt altitude hpopt search finds where ##EQU76## as shown in FIG. 11B.
In the higher dimension cases, where there are n variables, the following general method is employed for finding optimum altitude hpopt and maximum altitude for long range cruise speed MLRC, FIG. 8B, or optimum altitude hpopt for maximum speed Mmax at rated fuelflow Wfr, FIG. 8C. Gradients must be computed at each initial guess; the number of initial guesses exceeding the number of dimensions n by at least 1. The gradients are then used to fit a "second order surface" to the real surface on which the summit or other point of interest on the best fit surface is found. As the one dimensional case depends on the first derivative being approximately linear, this higher dimension method, depends on similar restrictions on the gradient, such that:
which implies that the surface itself may be expressed as:
f=1/2xT Ax+bT x+C=0 for A=AT
Having chosen initial guesses, xo, a "Matrix of Gradients" is expressed as: ##EQU77##
Solving for C yields
C=(MT M)-1 MT ∇Fo if m>n+1, or C=M-1 ∇VFo if m=n+1, A and b are then easily extracted from C. If the point of interest is the summit of the surface, this point can be found from:
x1 =-A-1 b
is the best estimate of where ∇f goes to zero.
Using this method, maximum speed Mmax and optimum altitude hpopt therefor, FIG. 8C, are determined. Assuming that pressure altitude hp and true airspeed VT are the only independent variables, such that weight W and temperature TDISA are completely specified by hp and fuelflow Wf is at the rated fuelflow value Wfr, then Wfr is specified given hp and VT.
Defining a residual climb thrust Trc as before, T-D as a function of hp and VT can be plotted in the form of a contour map as shown in FIG. 12. The point on the contour T-D=Trc which maximizes VT is the point at which maximum true airspeed occurs for any altitude subject to the constraint that the aircraft be able to initially maintain a rate of climb of 100 fpm. The coordinates of that point give the maximum airspeed Mmax and the altitude at which it will occur, the latter taken to be optimum altitude hpopt for maximum speed Mmax.
At any point on a given T-D contour, a tangent at that point is orthogonal to the gradient at that same point. Thus: ##EQU78##
The slope of the contours at a given point can then be expressed as: ##EQU79##
The method for finding maximum speed Mmax and optimum altitude hpopt arises from noting that ##EQU80## at maximum speed Mmax and optimum altitude hpopt. This condition manifests itself on the contour map as ##EQU81## which in conjunction with the fact that T-D=Trc gives the information required to move toward a final solution in both axes, simultaneously. The following steps are then performed:
1) Choose 3 initial guess points: xo.sbsb.a, xo.sbsb.b and xo.sbsb.c, as follows: ##EQU82## where hpoa =crossover altitude for Vmo /Mmo.
VT.sbsb.oa =maximum speed at hpoa (max speed, fixed alt case) ##EQU83## where ΔhP.sbsb.o =4000'
ΔVT.sbsb.o =0.05 mach expressed as true airspeed
If Mo.sbsb.a, the mach number corresponding to VT.sbsb.oa, is larger than Mmo, Mmax is taken to be Mmo and hpopt to be the crossover altitude and the algorithm is complete. Otherwise,
2) Compute the gradient of the T-D surface at the first initial point: ##EQU84## If ##EQU85## Vt.sbsb.max and hpopt are found by solving the following simultaneous equations: ##EQU86##
VT =VT.sbsb.1 +Vh.sbsb.1 (hp -hPoa)
VT.sbsb.1 =f(Vmo, hpmwr, TDISA) ##EQU87## f is the well known function relating true airspeed to calibrated airspeed, altitude and temperature.
Mmax is then taken from VT.sbsb.max and the algorithm is complete.
Otherwise ##EQU88## and the algorithm continues with step 3).
3) Assuming that (T-D) (x) can be approximately modeled as: ##EQU89##
Then the gradient is:
∇T(x)≈b+1/2(A+AT)Δx=b+AΔx since A=AT
A and b can then be estimated from the values computed in Steps 3 and 4, and ##EQU90## where ∇T(xo.sbsb.b) and ∇T(xo.sbsb.c) are completed in the same manner as ∇T(xo.sbsb.a).
and A and b are extracted from C.
4) The best estimate of the point x1 =[hP1 VT.sbsb.1 ]T (where VT.sbsb.1 is maximum speed and hp1 is optimum altitude for maximum speed) is found from the constraints: ##EQU91##
(T-D)(x1)=(T-D)(xo.sbsb.a)+bT (x1 -xo.sbsb.a)+1/2(x1 -xo.sbsb.a)T A(x1 -xo.sbsb.a)=0
Two simultaneous equations, one linear and one quadratic, in hP1 and VT.sbsb.1 can then be written. The solution for which ##EQU92## is correct.
5) The point x1 then becomes the new xo.sbsb.a. New guesses for xo.sbsb.b and xo.sbsb.c are chosen which have an order of magnitude less displacement from xo.sbsb.a than the displacement used on the first iteration. Steps 1-5 are then iterated until the movement in mach is less than 0.003 and the movement in altitude is less than 100 ft.
The long range cruise speed MLRC and optimum altitude hpopt and maximum altitude hpmax for MLRC, FIG. 8B, is calculated in a similar manner as shown below:
a) Maximum altitude hpmax is found first because if the optimum altitude hpopt is limited by the maximum altitude hpmax, then no further iterations are required to find optimum altitude hpopt.
b) A seed mach Mseed and altitude hp is chosen as an initial guess: hp =0.9 hceil where hceil is the ceiling altitude in ft. ##EQU93## where S is wing area in ft2 and δ is the atmospheric pressure ratio.
c) Based on this initial guess, SR and (T-D) surfaces are computed utilizing general surface search techniques as described previously with respect to determination of maximum speed. The T-D surface is computed using rated fuelflow as an input, and this therefore represents the maximum excess thrust at a given speed M and altitude hp and is shown in FIG. 14.
d) Long range cruise speed MLRC is defined to be where SR is 99% of the peak SR at the high speed side of the peak as shown in FIG. 13.
e) The maximum altitude hpmax for MLRC is the place where T-D=0 intercepts the 99% SR maximum line as shown by FIG. 14.
f) The maximum specific range SR.sbsb.max is given by the following two conditions: ##EQU94## Equation I is solved to yield VT as a function of hPo. Inserting VT into Equation II yields an expression for SR.sbsb.max as a function of hp.
g) The intercept on the T-D surface, representing hpmax, FIG. 14, satisfies the following two conditions: ##EQU95## By inserting the expression for SR.sbsb.max as a function of hp into Equation III, the problem of estimating hp and VT is reduced to a standard two-equations-two-unknowns computational problem.
h) Optimum altitude hpopt for MLRC is then found. If at the maximum altitude ##EQU96## then hpopt is set equal to hpmax. Otherwise, the hpopt is determined at the point where SR peaks with respect to altitude hp. This condition is written as: ##EQU97## Once again, using Equation III and V, the problem becomes one of two-unknowns-two-equations.
In many systems, it is difficult to say how accurate the predictions are after a fixed number of flights, which is dependent on the learning portion 12. This accuracy problem arises because the degree of accuracy depends on the kinds of flights that have been flown, i.e. an operator whose first ten flights have "visited" all comers of the flight envelope will have a fairly settled and accurate system, whereas an operator whose first ten flights were essentially duplicates of one another will have significant changes occur in the learning portion 12 when the operator finally operates in these corners of the flight envelope. The present invention, however, provides an observable indicator shown as figure of merit 24, FIG. 1, which indicates accuracy of the current estimate of fuel remaining at the destination. The figure of merit, FOM, is determined utilizing the covariance matrix P along with Weight WTOC, altitude hPtoc, speed MfTOC and fuelflow WfTOC, all at top of climb, as follows: ##EQU98## where HT is comprised of the first seven terms of the normalized (T-D) H matrix evaluated with the aforementioned top-of-climb terms,
PT is the upper left 7×7 portion of the (T-D) covariance P, and
tETE is estimated time enroute
Those skilled in the art will recognize that only preferred embodiments of the present invention have been disclosed herein, that other advantages may be found and realized, and that various modifications may be suggested by those versed in the art. It should be understood that the embodiment shown herein may be altered and modified without departing from the true spirit and scope of the invention as defined in the accompanying claims.
|Cited Patent||Filing date||Publication date||Applicant||Title|
|US4837695 *||Mar 12, 1986||Jun 6, 1989||E-Systems, Inc.||Method and apparatus for predicting and monitoring aircraft takeoff performance|
|US4843554 *||Aug 6, 1987||Jun 27, 1989||The United States Of America As Represented By The United States National Aeronautics And Space Administration||Airplane takeoff and landing performance monitoring system|
|US4868755 *||May 18, 1987||Sep 19, 1989||Texas Instruments Incorporated||Expert vehicle control system|
|US4980833 *||Jul 26, 1988||Dec 25, 1990||The University Of Tennessee Research Corporation||Airplane take-off monitor with learning feature|
|US5023797 *||Sep 20, 1990||Jun 11, 1991||United Technologies Corporation||Range maximizing, stable, helicopter cruise fuel conservation|
|US5070458 *||Mar 31, 1989||Dec 3, 1991||Honeywell Inc.||Method of analyzing and predicting both airplane and engine performance characteristics|
|1||Nelms, Jr. et al., "Preliminary Performance Estimates of an Oblique, All-Wing, Remotely piloted Vehicle for Air-to-Air Combat", NASA Ames Research Center, Moffett Field, CA, Jul. 1974, Abstract Only.|
|2||*||Nelms, Jr. et al., Preliminary Performance Estimates of an Oblique, All Wing, Remotely piloted Vehicle for Air to Air Combat , NASA Ames Research Center, Moffett Field, CA, Jul. 1974, Abstract Only.|
|3||*||Stevens et al., Aircraft Control and Simulation , John Wiley & Sons, Inc., NY, 1992, pp. 139 152.|
|4||Stevens et al., Aircraft Control and Simulation, John Wiley & Sons, Inc., NY, 1992, pp. 139-152.|
|Citing Patent||Filing date||Publication date||Applicant||Title|
|US5901272 *||Oct 24, 1996||May 4, 1999||The United States Of America As Represented By The Secretary Of The Navy||Neural network based helicopter low airspeed indicator|
|US6169496 *||Dec 9, 1998||Jan 2, 2001||Exigent International, Inc.||Banked flight stall warning device and method|
|US6390417 *||Jun 29, 2000||May 21, 2002||Honda Giken Kogyo Kabushiki Kaisha||Drag control system for flying machine, process for estimating drag of flying machine, boundary layer control system, and boundary layer control process|
|US6401056 *||Dec 27, 1999||Jun 4, 2002||General Electric Company||Methods and apparatus for evaluating tool performance|
|US6745114||Oct 17, 2002||Jun 1, 2004||Eurocopter||Process and device for determining in real time the behavior of a craft, in particular of an aircraft|
|US7177785 *||Aug 22, 2003||Feb 13, 2007||Honeywell International, Inc.||Systems and methods for improved aircraft performance predictions|
|US7296006 *||Feb 13, 2004||Nov 13, 2007||Simmonds Precision Products, Inc.||Method of inferring rotorcraft gross weight|
|US7853369 *||Nov 9, 2007||Dec 14, 2010||Airbus France||Active pitch control method and device for an aircraft|
|US8311687 *||Jul 30, 2010||Nov 13, 2012||Ge Aviation Systems Llc||Method and system for vertical navigation using time-of-arrival control|
|US8359542||Aug 13, 2010||Jan 22, 2013||Lockheed Martin Corporation||Machines, program products, and computer-implemented methods for interactive aircraft performance substantiation|
|US8509967 *||Aug 26, 2009||Aug 13, 2013||Thales||System and method for calculating flight predictions by vertical sections|
|US8596579 *||Oct 19, 2004||Dec 3, 2013||Airbus Operations Sas||Procedure and device for optimizing the quantity of fuel transferred on an aircraft during at least one in-fligh fuel transfer|
|US8768607 *||Sep 21, 2010||Jul 1, 2014||The Boeing Company||Managing fuel in aircraft|
|US9058040||Feb 27, 2009||Jun 16, 2015||The Boeing Company||Automatic pilot pitch angle compensation|
|US9085371 *||Nov 20, 2008||Jul 21, 2015||The Boeing Company||Automatic throttle roll angle compensation|
|US9248914||May 30, 2014||Feb 2, 2016||The Boeing Company||Managing fuel in aircraft|
|US20030065482 *||Dec 4, 2001||Apr 3, 2003||Simmonds Precision Products, Inc.||Method and apparatus for selecting condition indicators in determining the health of a component|
|US20040193386 *||Feb 13, 2004||Sep 30, 2004||Flynn Timothy D.||Method of inferring rotorcraft gross weight|
|US20050043934 *||Aug 22, 2003||Feb 24, 2005||Hartmann Gary L.||Intelligent database for performance predictions|
|US20050116109 *||Oct 19, 2004||Jun 2, 2005||Airbus France||Procedure and device for optimizing the quantity of fuel transferred on an aircraft during at least one in-fligh fuel transfer|
|US20100042270 *||Nov 9, 2007||Feb 18, 2010||Airbus France||Active pitch control method and device for an aircraft|
|US20100082185 *||Aug 26, 2009||Apr 1, 2010||Thales||System and Method for Calculating Flight Predictions by Vertical Sections|
|US20100125378 *||Nov 20, 2008||May 20, 2010||Frederick Charles Henry Blechen||Automatic throttle roll angle compensation|
|US20100222944 *||Feb 27, 2009||Sep 2, 2010||Frederick Charles Henry Blechen||Automatic pilot pitch angle compensation|
|US20120029735 *||Jul 30, 2010||Feb 2, 2012||Michael John Bakker||Method and system for vertical navigation using time-of-arrival control|
|US20120072098 *||Sep 21, 2010||Mar 22, 2012||The Boeing Company||Managing Fuel in Aircraft|
|US20120265374 *||Apr 15, 2011||Oct 18, 2012||Thomas Edward Yochum||Aircraft vertical trajectory optimization method|
|US20140172619 *||Dec 14, 2012||Jun 19, 2014||Bell Helicopter Textron Inc.||Aircraft Sales Tool|
|USRE41381 *||Jun 12, 2003||Jun 22, 2010||Stabile James R||Method of calculating oxygen required and system for monitoring oxygen supply and/or calculating flight level after emergency cabin decompression|
|EP1312998A1 *||Oct 1, 2002||May 21, 2003||Eurocopter||Method and device for real-time determining the behaviour of a vehicle, in particular of an aircraft|
|WO2005019955A1 *||Aug 20, 2004||Mar 3, 2005||Honeywell International Inc.||Intelligent database for performance predictions|
|U.S. Classification||701/99, 701/15, 701/16|
|Jul 31, 2000||FPAY||Fee payment|
Year of fee payment: 4
|Jun 10, 2003||RR||Request for reexamination filed|
Effective date: 20030514
|Jun 29, 2004||FPAY||Fee payment|
Year of fee payment: 8
|Jul 17, 2008||FPAY||Fee payment|
Year of fee payment: 12